Content
# sqlite-reader-mcp
MCP server for reading SQLite databases.
## Overview
`sqlite-reader-mcp` is a lightweight MCP server that provides read-only access to SQLite databases. It allows users to execute SELECT queries, list tables, and describe table schemas.
## Features
- **Read-Only Access**: Ensures data integrity by only allowing read operations (SELECT queries).
- **Secure Path Handling**: Restricts database access to pre-approved absolute file paths and directories.
- **SQL Validation**: Parses and validates SQL queries, permitting only `SELECT` and `WITH` statements to prevent unintended operations.
- **Asynchronous Operations**: Utilizes `aiosqlite` for non-blocking database interactions.
- **Row Limiting**: Automatically limits the number of rows returned by queries to prevent excessive data retrieval.
- **MCP Tools**:
- `read_query`: Execute a SELECT SQL query on a specified SQLite database.
- `list_tables`: List all tables within a given SQLite database.
- `describe_table`: Get schema information (column names, types, etc.) for a specific table.
## Requirements
- Python 3.13+
- `aiosqlite >= 0.21.0`
- `fastmcp >= 2.6.1`
- `sqlparse >= 0.5.3`
## Installation
The recommended installation method is with `uv`:
```bash
uv tool install git+https://github.com/abhinavnatarajan/sqlite-reader-mcp
```
## Usage
```bash
sqlite-reader-mcp --paths /path/to/your/database.db /path/to/another/allowed_directory
```
**Arguments:**
- `--paths` (or `-p`): A list of absolute paths to SQLite database files or directories containing them that the server is allowed to access.
Once the server is running, you can interact with it using a FastMCP client, utilizing the following tools:
### `read_query`
Executes a SELECT query on a specified SQLite database file.
- **Args**:
- `file_path` (str): Absolute path to the SQLite database file.
- `query` (str): The SELECT SQL query to execute.
- `params` (Optional[List[Any]]): Optional list of parameters for the query.
- `fetch_all` (bool): If `True` (default), fetches all results. If `False`, fetches one row.
- `row_limit` (int): Maximum number of rows to return (default 1000).
- **Returns**: A list of dictionaries, where each dictionary represents a row from the query result.
### `list_tables`
Lists all tables in the specified SQLite database file.
- **Args**:
- `file_path` (str): Absolute path to the SQLite database file.
- **Returns**: A list of table names (strings).
### `describe_table`
Provides detailed schema information for a specific table in a SQLite database.
- **Args**:
- `file_path` (str): Absolute path to the SQLite database file.
- `table_name` (str): The name of the table to describe.
- **Returns**: A list of dictionaries, each describing a column (e.g., name, type, notnull, default value, primary key status).
## Security
- **Read-Only**: The server is strictly read-only. No `INSERT`, `UPDATE`, `DELETE`, or other data modification SQL commands are permitted.
- **Path Whitelisting**: Access is restricted to SQLite files located in paths explicitly provided via the `--paths` command-line argument. Paths must be absolute and existent.
- **Query Validation**: All SQL queries are parsed. Only `SELECT` statements (including those with Common Table Expressions using `WITH`) are allowed. Multiple statements in a single query string are rejected. Trailing semicolons are automatically removed.
## Development
This project is built using `setuptools`. Configuration is in `pyproject.toml`.
Connection Info
You Might Also Like
markitdown
MarkItDown-MCP is a lightweight server for converting URIs to Markdown.
markitdown
Python tool for converting files and office documents to Markdown.
Filesystem
Node.js MCP Server for filesystem operations with dynamic access control.
TrendRadar
TrendRadar: Your hotspot assistant for real news in just 30 seconds.
mempalace
The highest-scoring AI memory system ever benchmarked. And it's free.
mempalace
The highest-scoring AI memory system ever benchmarked. And it's free.